A Multi-Weight-Driven Intelligent Inspection Route Optimization Method for Urban Roads

Abstract

This application discloses a multi-weight-driven intelligent inspection route optimization method for urban roads. The method includes acquiring several initial records of road defects, associating these records using a spatiotemporal matching algorithm, and generating a defect evolution chain based on time sequence. Basic vulnerability features are extracted from the defect evolution chain, and internal vulnerability scores and external influence scores are calculated based on these features. These scores are then weighted to obtain a comprehensive vulnerability index, which is mapped to inspection decision weights. Based on these weights, tasks are aggregated for the road segments to be inspected, resulting in inspection task packages and corresponding aggregation weights. A multi-vehicle path planning model is constructed, and the inspection task packages and aggregation weights are input into the model to obtain an initial inspection plan. This application can extract defect recurrence patterns from historical maintenance records, improving the probability of discovering high-risk defect points and the scientific rigor of inspection plans.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to an intelligent inspection route optimization method for urban roads driven by multiple weights. Background Technology

[0002] With the acceleration of urbanization, the scale of urban road networks is growing exponentially. Traditional "manual inspection" or "fixed-period inspection" models can no longer meet the growing demand for refined management and maintenance. At present, road asset management is transforming from "passive repair" to "preventive maintenance," and using vehicle-mounted mobile measurement systems for full-coverage inspection has become the mainstream method.

[0003] However, given the vast road network and limited inspection vehicle resources, developing efficient inspection plans has become a major challenge for the industry. Existing inspection scheduling is mostly based on administrative divisions or simple geometric distances, lacking consideration for the actual "health condition" and "urgency" of road sections. This "average effort" model results in insufficient inspection frequency for high-risk sections, while healthy sections are over-inspected, leading to a double waste of computing and transportation capacity.

[0004] Against this backdrop, the key to improving the efficiency of urban road operation and maintenance lies in how to construct a multi-dimensional road segment vulnerability evaluation system based on massive historical maintenance records and the spatiotemporal evolution patterns of road damage, and in order to achieve intelligent path planning under multiple constraints. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent inspection route optimization method for urban roads driven by multiple weights, in order to solve the technical problems of low inspection efficiency and inability to respond to complex environmental factors in existing inspection route planning methods. The various technical effects of the preferred technical solutions provided in this application are detailed below.

[0006] To achieve the above objectives, this application provides the following technical solution: This application provides a multi-weight-driven intelligent inspection route optimization method for urban roads, comprising: acquiring several initial records of road defects; associating the initial records of road defects using a spatiotemporal matching algorithm to generate a disease evolution chain based on time sequence; obtaining basic vulnerability features from the disease evolution chain; calculating internal vulnerability score and external influence score based on the basic vulnerability features; weighting the internal vulnerability score and the external influence score to obtain a comprehensive vulnerability index; mapping the comprehensive vulnerability index to inspection decision weights; aggregating tasks for the road segments to be inspected based on the inspection decision weights to obtain inspection task packages and corresponding aggregation weights; constructing a multi-vehicle path planning model; inputting the inspection task packages and the aggregation weights into the multi-vehicle path planning model to obtain an initial inspection scheme, wherein the multi-vehicle path planning model uses the selection of the inspection task package with the largest aggregation weight and the minimization of inspection cost and inspection time as objective functions.

[0007] In some embodiments, the step of using a spatiotemporal matching algorithm to associate the initial disease records and generate a disease evolution chain includes: selecting one of the initial disease records as a seed record; searching within a preset road segment range based on the seed record to obtain multiple candidate records, wherein the candidate records include initial disease records and maintenance records in subsequent time steps; performing spatial and temporal consistency matching on the candidate records; determining the association between the candidate records that pass the spatial and temporal consistency matching and the seed record to obtain repair nodes and recurrence nodes; and sequentially connecting the discovery node, upgrade node, repair node, and recurrence node corresponding to the same disease location in chronological order to obtain the disease evolution chain.

[0008] In some embodiments, the step of aggregating tasks based on the inspection decision weights of the road segments to be inspected to obtain inspection task packages and corresponding aggregation weights includes: constructing a road segment graph model based on the road segments to be inspected; judging and summarizing each adjacent road segment in the road segment graph model according to preset conditions to obtain a candidate aggregation set, wherein the preset conditions include at least the difference in inspection decision weights of adjacent road segments not exceeding a first preset threshold; performing structured encoding processing on the candidate aggregation set to obtain structured input data; performing aggregation judgment on the structured input data through a large language model to output a task package label corresponding to each road segment to be inspected; merging multiple road segments to be inspected into the inspection task package based on the task package label; and simultaneously calculating the aggregation weight of the inspection task package, wherein the aggregation weight is used to characterize the node reward value of the vehicle executing the inspection task package.

[0009] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads further includes: acquiring inspection information of each inspection vehicle in real time, and performing state freezing processing on some of the inspection task packages based on the inspection information.

[0010] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads further includes: receiving dynamic monitoring data; dynamically updating the unlocked inspection task packages and the aggregated weights based on the dynamic monitoring data to obtain updated task packages and updated aggregated weights; inputting the newly added task packages, the updated task packages, and the corresponding aggregated weights into the multi-vehicle path planning model to perform local re-optimization on the remaining inspection routes; and outputting an updated inspection plan based on the results of the local re-optimization to update the initial inspection plan.

[0011] In some embodiments, the local re-optimization of the remaining inspection path includes: traversing any two inspection task packages in the remaining inspection path, obtaining several feasible insertion positions between any two inspection task packages that satisfy the constraints; calculating the change in the objective function of the feasible insertion positions, and selecting the feasible insertion position with the smallest change in the objective function to insert the new task package.

[0012] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads further includes: performing partial exchange of unexecuted task packages in different vehicle paths, wherein the partial exchange includes at least exchanging the access order of two unexecuted task packages within the same inspection path, and / or transferring unexecuted task packages in one vehicle path to another vehicle path, and / or exchanging unexecuted task packages between two vehicles.

[0013] In some embodiments, the aggregation weights are calculated as follows: Among them, ΣWtask i η1 is the cumulative value of the inspection decision weights of all the road sections to be inspected within the inspection task package; Uzone is the timeliness urgency of the inspection task package; Czone is the spatial compactness of the inspection task package; Dzone is the additional detour cost or dispersion of the inspection task package; η1 is the first preset weight coefficient, η2 is the second preset weight coefficient, η3 is the third preset weight coefficient, and η4 is the fourth preset weight coefficient.

[0014] In some embodiments, the objective function is expressed by the following formula: Among them, W zone_p y represents the aggregate weight of the inspection task package p; p Indicates whether the inspection task package p is executed; cpq represents the travel cost from inspection task package p to inspection task package q; T kλ1 represents the total inspection time of vehicle k; λ2 is the first preset adjustment coefficient, λ3 is the second preset adjustment coefficient, and λ4 is the third preset adjustment coefficient.

[0015] In some embodiments, the change in the objective function is expressed as: Where cpu represents the travel cost from the inspection task package p to the new task package u, cuq represents the travel cost from the new task package u to the inspection task package q; cpq represents the travel cost from the inspection task package p directly to the inspection task package q; ΔT represents the additional delay on the vehicle's remaining path after inserting the new task package; W zoneu The aggregate weight of the newly added task package is represented by α, β, and γ.

[0016] Implementing one of the above-mentioned technical solutions of this application has the following advantages or beneficial effects: In this application, firstly, the disease evolution chain is obtained based on the spatiotemporal matching algorithm, then the basic vulnerability characteristics are obtained from the disease evolution chain to calculate the inspection decision weight, and the road segments to be inspected are aggregated into inspection task packages and corresponding aggregation weights based on the inspection decision weights. Finally, the initial inspection plan is output based on the inspection task packages and corresponding aggregation weights through the multi-vehicle path planning model.

[0017] In this context, by acquiring time-series-based disease evolution chains, it is possible to extract disease recurrence patterns from historical maintenance records, enabling inspection resources to be prioritized for high-risk road sections. This improves the probability of discovering high-risk disease points and the scientific rigor of inspection plans. Furthermore, aggregating road sections to be inspected into inspection task packages based on inspection decision weights allows subsequent inspection plans to consider risk priority, traffic conditions, and operational timeliness, thereby improving the coverage efficiency of high-value inspection tasks. Simultaneously, compared to solving discrete road sections one by one, this application reduces the number of path planning nodes and the scale of combinatorial optimization, thus improving the solution efficiency and engineering practicality of multi-vehicle path planning models in large-scale urban road networks. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a flowchart illustrating the intelligent inspection route optimization method for urban roads driven by multiple weights, according to an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, various exemplary embodiments described below will be referenced to the accompanying drawings, which form part of the exemplary embodiments and depict various exemplary embodiments that may be adopted to implement this application. Unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. It should be understood that they are merely examples of processes, methods, and apparatuses consistent with some aspects of this application disclosed as detailed in the appended claims, and other embodiments may be used, or structural and functional modifications may be made to the embodiments listed herein without departing from the scope and spirit of this application.

[0020] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," etc., indicate the orientation or positional relationship based on the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the referred element must have a specific orientation, or be constructed and operated in a specific orientation. The terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. The term "multiple" means two or more. The terms "connected" and "linked" should be interpreted broadly, for example, they can be fixed connections, detachable connections, integral connections, mechanical connections, electrical connections, communication connections, direct connections, indirect connections through an intermediate medium, and can be the internal connection of two elements or the interaction relationship between two elements. The term "and / or" includes any and all combinations of one or more of the related listed items. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0021] To illustrate the technical solutions described in this application, specific embodiments are provided below, showing only the parts related to the embodiments of this application.

[0022] like Figure 1 As shown, this application provides a multi-weight-driven intelligent inspection route optimization method for urban roads, including the following steps (steps S1 to S3): S1. Obtain several initial disease records, use a spatiotemporal matching algorithm to associate the initial disease records, and generate a disease evolution chain based on time sequence.

[0023] In some embodiments, obtaining several initial records of road defects may include: extracting defect data based on a historical road maintenance database, standardizing the defect data, and obtaining initial records of road defects. The defect data in this application embodiment may refer to road defect data, that is, a collection of information describing various "defects" or "damage" states of the road.

[0024] Specifically, the damage data can include damage records, maintenance records, and inspection records. Standardization processing can include mapping and unifying the formats of road names, road numbers, station numbers, latitude and longitude coordinates, lane numbers, damage types, damage levels, recording times, and maintenance status from different sources of damage data.

[0025] Furthermore, for damage records with latitude and longitude coordinates, they can be projected onto the corresponding road centerline through map matching and converted into a unified linear reference position identifier; for damage data with location drift, they can be corrected by adjacent inspection trajectories and road topology to ensure that damage records at different times are comparable under the same spatial reference system.

[0026] After standardizing the disease data, each initial disease record can be used as a seed node. Subsequent disease records and maintenance records can be retrieved in chronological order and associated using a spatiotemporal matching algorithm to generate a disease evolution chain.

[0027] In some embodiments, using a spatiotemporal matching algorithm to associate initial disease records and generate a disease evolution chain may include: selecting any initial disease record as a seed record, searching within a preset road segment range based on the seed record to obtain multiple candidate records, wherein the candidate records include disease records and maintenance records in subsequent time steps; performing spatial and temporal consistency matching on the candidate records, determining the association between the candidate records that pass the spatial and temporal consistency matching and the seed record, obtaining repair nodes and recurrence nodes, and sequentially connecting the discovery node, upgrade node, repair node, and recurrence node corresponding to the same disease location in chronological order to obtain a disease evolution chain.

[0028] The disease evolution chain includes a unique disease evolution chain ID. The disease evolution chain can be used to output the disease type, initial discovery time, repair time, recurrence time, disease severity change process, road location, and associated confidence level.

[0029] Specifically, based on seed records, within a preset road segment range that is the same as or topologically connected to the seed record, defect records and maintenance records within subsequent time windows can be retrieved to form a candidate record set. Each candidate record can satisfy at least one spatial constraint condition, which may include having the same spatial road number, a linear reference position difference not exceeding a first preset distance threshold, or a road topological adjacency relationship satisfying continuous passage conditions.

[0030] Spatial consistency matching can refer to calculating the spatial deviation between the seed record and the candidate record. The spatial deviation can include location distance, lane consistency, and driving direction consistency. Further, when the candidate record and the seed record are located on the same road, in the same lane, or in adjacent permissible lanes, and the location distance is less than a first preset distance threshold, it can be determined that the spatial consistency matching has passed. Among them, the location distance is preferably the projection distance along the centerline of the road to reduce the error caused by GPS drift.

[0031] For candidate records that pass spatial consistency matching, the time interval between them and the seed record is calculated. If the time interval is within a preset range, or the occurrence time of the candidate record is within a preset time window, it can be determined that it has passed time consistency matching. In the maintenance event identification scenario, the maintenance time should be later than the discovery time; in the re-damage identification scenario, the re-occurrence time should be later than the maintenance time.

[0032] For candidate records that pass both spatial and temporal matching, the disease type, disease level, and disease morphological characteristics are further compared. If the disease types are the same, or belong to a preset combination of evolvable disease categories, and the disease level changes conform to preset evolution rules, then the candidate record and the seed record are determined to belong to the same disease evolution process, which is considered a disease attribute association between the candidate record and the seed record. The evolvable disease category combination can include sequential association types such as crack expansion, pit recurrence, and subsidence aggravation.

[0033] When a repair record matching the current disease record is found, it can be designated as a repair node for that disease stage. Following the repair node, the search continues within subsequent time windows for disease records that meet the spatiotemporal matching criteria. If a matching record exists, it is marked as a recurrence node for that location, establishing a chain-like relationship of "discovery—evolution—repair—re-damage." Repair nodes correspond to repair time, and correspondingly, recurrence nodes correspond to recurrence time.

[0034] The discovery node, escalation node, repair node, and recurrence node corresponding to the same disease location are sequentially linked in chronological order to generate a unique disease evolution chain ID. The chain is then output, including the disease type, initial discovery time, repair time, recurrence time, disease severity change process, road location, and association confidence level. Furthermore, if no recurrence of the disease is detected within the current observation period, the disease evolution chain can be marked as non-recurrence.

[0035] In some embodiments, the re-damage interval can be the time difference between the repair time and the first recurrence time of the first match. Specifically, for each disease evolution chain, the repair time and the first recurrence time of the first match after the repair node can be extracted, and the time difference between the two can be calculated as the re-damage interval at that disease location.

[0036] In some embodiments, the re-damage interval can be expressed as: Where Trecur is the recurrence time and Trepair is the repair time.

[0037] If no matching recurrence node is detected within the current observation period, the time difference between the repair time and the current time is recorded as the current non-recurrence interval, which is used as the observed re-damage interval for that location. Preferably, such disease evolution chains can be marked as non-recurrence or censored states to distinguish them from disease evolution chains that have already recurred. In other embodiments, the intelligent inspection route optimization method for urban roads driven by multiple weights may further include: using a survival analysis model to estimate the risk of disease recurrence to obtain the probability of disease recurrence or the expected recurrence time in the future.

[0038] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may also include using a graph database-based association query method to construct a time-series graph of disease nodes, maintenance nodes and time relationships, so as to realize the tracking and retrieval of the relationship of the entire life cycle of the disease.

[0039] This application constructs a disease lifecycle evolution chain that includes discovery, evolution, repair, and recurrence nodes, and calculates the re-damage interval based on the repair time and recurrence time of the same disease location, thereby extracting disease recurrence patterns from historical maintenance records.

[0040] Compared to existing technologies that rely solely on road grade or fixed frequency for inspections, this application can identify locations of persistent defects with high recurrence rates and rapid deterioration. This allows inspection resources to be prioritized for high-risk road sections, reducing redundant inspections of low-risk, healthy sections and increasing the probability of detecting high-risk defects and the scientific rigor of inspection plans. This application achieves recurrence risk identification based on the entire lifecycle evolution chain of defects, improving the targeting and accuracy of inspection decisions.

[0041] In some embodiments, the rate of change of disease level over time can be calculated based on the changes in disease level of adjacent disease nodes in the disease evolution chain and their corresponding time intervals, and this rate can be used as the disease evolution rate.

[0042] In some embodiments, the disease evolution rate v can be expressed as: Where Δg is the change in disease level between two adjacent disease nodes within the same disease evolution chain, and Δt is the corresponding time interval.

[0043] For a disease evolution chain that includes multiple disease evolution stages, the average, maximum, or weighted average of the evolution rates of each stage can be taken as the final evolution rate at that disease location to reflect the speed of disease deterioration.

[0044] In some embodiments, the number of repeated repairs at the same location within a preset observation period can be counted based on the disease evolution chain and historical maintenance records to obtain the historical number of repeated repairs.

[0045] The preset observation period can be set according to actual business needs, such as six months, one year, or several inspection cycles. The more times the repair is repeated, the higher the structural vulnerability or risk of stubborn defects at that location.

[0046] In some embodiments, basic road attributes may include at least one of the following: road grade, number of lanes, road function type, design life, or historical traffic load. Road grade includes arterial roads, secondary arterial roads, local roads, or expressways, and may be mapped to corresponding grade coefficients to characterize differences in inspection priority and vulnerability sensitivity among different roads.

[0047] S2. Obtain basic vulnerability characteristics from the disease evolution chain, calculate the internal vulnerability score and external influence score based on the basic vulnerability characteristics, perform weighted processing on the internal vulnerability score and external influence score to obtain the comprehensive vulnerability index, map the comprehensive vulnerability index into the inspection decision weight, and aggregate the tasks of the road section to be inspected based on the inspection decision weight to obtain the inspection task package and the corresponding aggregation weight.

[0048] Among them, basic vulnerability characteristics can include re-damage interval, disease evolution rate, number of historical repeated repairs, and basic road properties. Basic vulnerability characteristics are used to characterize the disease recurrence characteristics and physical vulnerability of the corresponding road segment.

[0049] In some embodiments, a time decay factor can be calculated based on the re-damage interval, and then combined with the disease evolution rate and the number of historical repeated repairs to obtain an intrinsic vulnerability score, which can be used to quantify the physical vulnerability of the road segment itself.

[0050] In some embodiments, the intrinsic vulnerability factor can be expressed by the following formula: Where Te is the re-damage interval, ln(T) e +e) represents the logarithmic function used to smooth the time interval and take its reciprocal, indicating that the shorter the interval, the higher the vulnerability; v is the disease evolution rate, characterizing how quickly the disease severity changes over time; n is the historical number of repeated repairs at this disease location, calculated using (1−e) −n The number of repairs is mapped to the interval [0, 1), indicating that the more repairs are done, the more vulnerable the road segment is; k1 is the first intrinsic factor weight coefficient, k2 is the second intrinsic factor weight coefficient, and k3 is the third intrinsic factor weight coefficient.

[0051] In some embodiments, the external factor impact score can be calculated by obtaining the current real-time traffic congestion index, the sensitivity of the surrounding environment (e.g., whether it is near high-traffic areas such as schools, hospitals, and transportation hubs), the inspection timeliness requirements, and the current density of diseases in the area.

[0052] In some embodiments, the external factor influence score can be expressed by the following formula: Where S is the sensitivity coefficient of the surrounding environment, which takes a higher value (e.g., 1.5) when the road segment is near a school, hospital or transportation hub, otherwise takes a baseline value (e.g., 1.0); ρ is the density of defects in the current road segment; R is the inspection time requirement, the shorter the remaining time, the larger the R value; C is the real-time traffic congestion index, which is used as the denominator to suppress the operation weight during congestion periods and realize off-peak inspection; λ1 is the first external factor weight coefficient, λ2 is the second external factor weight coefficient, λ3 is the third external factor weight coefficient, and μ is the congestion adjustment factor.

[0053] A comprehensive vulnerability index can be obtained by fusing the internal vulnerability score and the external impact score using a dynamic weighted model. The comprehensive vulnerability index can be expressed by the following formula: Where α is the preset weight of the internal vulnerability score, β is the preset weight of the external influence score, and α+β=1; Δt is the time interval between the last effective inspection of the road segment; γ is the time urgency factor.

[0054] The above formula integrates the risk of the disease itself and environmental constraints, and also utilizes the exponential term e. γ Δt The potential risks were simulated to increase non-linearly with the increase of inspection gaps, thereby generating a comprehensive vulnerability index that reflects the actual urgency of inspections on the current road section.

[0055] The comprehensive vulnerability index of this application can not only reflect the risk of the disease itself, but also the current working environment and the urgency of inspection. Compared with the method of planning inspection routes based solely on geometric distance or static level, it can more realistically represent the priority of road inspection.

[0056] In other embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may further include: using a fuzzy logic reasoning model, a rule-based scoring model, or a supervised learning regression model to predict the comprehensive vulnerability index, so as to achieve a quantitative assessment of the urgency of road section inspection.

[0057] The comprehensive vulnerability index of each road segment to be inspected is mapped to an inspection decision weight for route planning. The inspection decision weight is used to characterize the inspection priority, risk urgency, and resource investment value of the corresponding road segment at the current moment.

[0058] Preferably, the inspection decision weight can be obtained from the comprehensive vulnerability index after normalization, or it can be obtained from the comprehensive vulnerability index through a preset mapping function, so as to ensure that the weights between different road sections are comparable.

[0059] This application maps the comprehensive vulnerability index to inspection decision weights, enabling risk assessment results to be directly converted into calculable benefit values ​​in subsequent route planning and task scheduling. By establishing a unified mapping relationship from risk indicators to scheduling weights, effective connection can be achieved between the results of defect risk identification and inspection execution strategies, allowing high-risk and high-urgency road sections to receive higher priority during scheduling.

[0060] After obtaining the inspection decision weights for each road segment to be inspected, a road segment map model can be constructed by combining the topology of the urban road network. The road segment map model can be represented by the following formula: G=(V,E), In this diagram, node V represents the road segment to be inspected, and edge E represents a spatial adjacency between two road segments that allows for continuous passage. A spatial adjacency relationship can include at least one of the following: identical road numbers, road centerline projection distance less than a second preset distance threshold, being in the same area, or having a direct connection in the road network topology.

[0061] Furthermore, a structured feature vector can be extracted for each road segment to be inspected. The structured feature vector can include at least the road segment identifier, start and end coordinates, road grade, lane information, inspection decision weight, estimated inspection duration, allowable operation time window, distribution of damage types, sensitivity to the surrounding environment, and connectivity with adjacent road segments.

[0062] In some embodiments, task aggregation is performed on the road segment to be inspected based on the inspection decision weight to obtain an inspection task package and the corresponding aggregation weight, which may include: Based on the road segment to be inspected, a road segment map model is constructed. According to the preset conditions, each adjacent road segment in the road segment map model is judged and summarized to obtain a candidate aggregation set. The preset conditions include at least the difference between the inspection decision weights of adjacent road segments is not greater than the first preset threshold. The candidate aggregation set is structured and encoded to obtain structured input data. The structured input data is aggregated and judged by a large language model, and the task package label corresponding to each road segment to be inspected is output. Based on the task package label, multiple road segments to be inspected are merged into an inspection task package. At the same time, the aggregation weight of the inspection task package is calculated. The aggregation weight is used to represent the node reward value of the vehicle executing the inspection task package.

[0063] Specifically, based on the road segment graph model, starting from each high-weight road segment, we can search for adjacent and continuously passable surrounding road segments in a graph traversal manner to form a candidate aggregation set.

[0064] In some embodiments, the preset conditions may also include the distance between adjacent road segments on the centerline of the road not being greater than a second preset threshold, the allowable operation time windows of adjacent road segments overlapping, and adjacent road segments having the same or similar types of defects, operation objectives, or environmental constraints.

[0065] Using the above method, candidate connected subgraphs with continuous operation conditions can be selected first by filtering the road network topology and rule constraints, providing initial objects for subsequent intelligent aggregation.

[0066] The candidate aggregation set is subjected to structured encoding to obtain structured input data, which may include encoding the road segment structured features, adjacency relationships between road segments, weight distribution, time window constraints and environmental attributes in each candidate aggregation set into unified structured input data.

[0067] Structured input data can include at least the set of candidate road segments and their corresponding road segment numbers, the spatial location of each candidate road segment, their adjacency relationships and connectivity paths, the inspection decision weights for each candidate road segment, the time windows for each candidate road segment, the estimated operation time and inspection constraints, the types of defects, environmental sensitivity, and other auxiliary attributes for each candidate road segment. A candidate road segment can refer to a road segment to be inspected.

[0068] The structured input data is input into the large language model according to the preset prompt template, so that the large language model can combine spatial continuity, weight similarity, task timeliness and inspection target consistency to perform semantic merging judgment on candidate road segments.

[0069] The large language model, based on structured input data, performs aggregation judgment on each road segment in the candidate aggregation set and outputs the task package label to which each road segment belongs. When multiple road segments meet the conditions of spatial continuity, similar weights, compatible operation time windows, and consistent inspection targets, they are determined to be grouped into the same inspection task package.

[0070] Among them, consistent inspection targets can refer to multiple road sections being used for key re-inspection of similar diseases, key risk investigation, priority inspection around schools and hospitals, or road inspection tasks under the same time-efficiency level.

[0071] By introducing a large language model, we can further integrate semantic-level task similarity and task rationality on the basis of traditional topological adjacency and numerical threshold judgment, thereby avoiding the problems of task fragmentation or unreasonable task boundaries caused by simply relying on geometric distance clustering.

[0072] In other embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may also include: using a clustering algorithm based on geographical distance and weight similarity, or using a graph partitioning method based on road network graph structure, to aggregate spatially adjacent road segments with the same operational attributes.

[0073] In some embodiments, an inspection task package may include at least a task package number, a set of road segment numbers included in the inspection task package, the center location or representative location of the inspection task package, the total inspection length of the inspection task package, the estimated inspection operation time of the inspection task package, the unified operation time window of the inspection task package, the boundary information of the inspection task package and its internal connectivity, and the status information of the inspection task package. The status information of the inspection task package may include at least an unexecuted state, an in-process state, a completed state, and a locked state.

[0074] By grouping discrete road segments to be inspected into highly cohesive inspection task packages, the subsequent route planning object can be upgraded from the single road segment level to the area task level, thereby reducing the scale of route planning, reducing vehicle detours, and improving inspection continuity and scheduling efficiency.

[0075] For each generated inspection task package, the corresponding aggregate weight is calculated based on the inspection decision weight of each road segment within it, the time urgency of the task package, the spatial compactness, and the additional detour cost.

[0076] In some embodiments, the aggregation weights may be calculated as follows: Among them, ΣWtask i η1 is the cumulative value of the inspection decision weights of all road segments to be inspected within the inspection task package, used to reflect the overall inspection value of the inspection task package; Uzone is the timeliness urgency of the inspection task package, used to reflect the priority brought by the remaining executable time or the earliest deadline of each road segment within the inspection task package; Czone is the spatial compactness of the inspection task package, used to reflect the degree of spatial continuity between road segments within the inspection task package. The more compact the space, the more suitable it is for continuous inspection in one go; Dzone is the additional detour cost or dispersion of the inspection task package, used to reflect the additional path cost required to execute the inspection task package; η1 is the first preset weight coefficient, η2 is the second preset weight coefficient, η3 is the third preset weight coefficient, and η4 is the fourth preset weight coefficient.

[0077] After the above processing, each inspection task package corresponds to an aggregate weight, which serves as the node revenue value in subsequent multi-vehicle path planning.

[0078] This application integrates intrinsic characteristics such as re-damage interval, disease evolution rate, and number of repeated repairs with extrinsic characteristics such as real-time traffic congestion, environmental sensitivity, task timeliness requirements, and disease density to calculate a comprehensive vulnerability index, and further maps it to generate inspection decision weights, inspection task packages, and aggregation weights.

[0079] Therefore, the generated inspection routes no longer only aim at the shortest geometric distance, but can take into account risk priority, traffic conditions and operational timeliness. Under the same number of vehicles and operation time, it can improve the coverage efficiency of high-value inspection tasks and reduce the proportion of ineffective detours and low-value inspections.

[0080] This application first transforms discrete road segments to be inspected into inspection task packages with unified boundaries, service durations, and time window attributes. Then, it uses the inspection task packages as nodes to perform multi-vehicle route optimization. Compared with the existing technology that directly solves for massive discrete road segments one by one, this can reduce the number of path planning nodes, reduce the scale of combinatorial optimization, and improve the solution efficiency and engineering practicality of the scheduling model in large-scale urban road networks.

[0081] This application constructs a multi-factor driven comprehensive vulnerability evaluation mechanism, which improves the timeliness, rationality and adaptability of inspection route generation, reduces the solution complexity of path planning problems, and improves the feasibility of actual scheduling.

[0082] S3. Construct a multi-vehicle path planning model. Input the inspection task package and the aggregation weight into the multi-vehicle path planning model to obtain the initial inspection plan. The multi-vehicle path planning model takes selecting the task package with the largest aggregation weight and minimizing the inspection cost and inspection time as its objective functions.

[0083] Specifically, each inspection task package can be used as a node to be visited to construct a multi-vehicle routing model with time window constraints. The multi-vehicle routing model can be used to generate inspection solutions that balance maximizing task benefits and minimizing operating costs under conditions of limited vehicle resources, limited operation time, and dynamic changes in the road network.

[0084] In some embodiments, the input to the multi-vehicle routing model may also include a travel distance matrix or travel time matrix, estimated service duration, allowed operation time window, vehicle set, vehicle information, and constraints. These can be input into the multi-vehicle routing model along with inspection task packages and aggregated weights to obtain a more accurate initial inspection plan.

[0085] Specifically, vehicle information may include the starting location of each vehicle, remaining operating time, remaining range or other resource constraints, and constraints may include locked tasks, restricted areas or traffic restrictions, etc.

[0086] Furthermore, decision variables x(k, p, q) can be set to indicate whether vehicle k travels from inspection task package p to inspection task package q, y_p can be set to indicate whether inspection task package p is executed by any vehicle, and t(k, p) can be set to indicate the time when vehicle k arrives at inspection task package p.

[0087] Based on this, a joint optimization objective is established to maximize the total revenue of the executed task packages while minimizing the total vehicle travel cost and total operation time cost, under the premise of satisfying vehicle capacity constraints, path continuity constraints, and time window constraints.

[0088] In some embodiments, the objective function can be expressed by the following formula: Among them, W zone_p y represents the aggregate weight of the inspection task package p; p Indicates whether inspection task package p is executed; cpq represents the travel cost from inspection task package p to inspection task package q; T k λ1 represents the total inspection time of vehicle k; λ2 is the first preset adjustment coefficient, λ3 is the second preset adjustment coefficient, and λ4 is the third preset adjustment coefficient.

[0089] Maximizing the total task revenue in the objective function mentioned above can mean prioritizing the execution of task packages with higher aggregation weights under the conditions of limited inspection resources, number of vehicles, and operation time, so as to ensure that limited resources are prioritized to cover high-risk, high-urgency, and high-value inspection tasks, rather than simply pursuing the shortest total distance.

[0090] In some embodiments, the constraints of the multi-vehicle routing model may include each task package being executed by at most one vehicle, each vehicle starting from its corresponding starting point and accessing several inspection task packages according to the planned route, the arrival time of each inspection task package being within the allowed operation time window of the corresponding task package, the total travel time, total service time or total energy consumption of each vehicle on the execution route not exceeding the vehicle's remaining resource limit, inspection task packages that are already locked not being reallocated or having their access order changed, and each vehicle route needing to meet at least one of the conditions of road network accessibility and continuous passage.

[0091] The multi-vehicle routing model can automatically generate an initial inspection plan based on inspection task packages and their aggregate weights, combined with vehicle resources and time window constraints, to determine which inspection task packages each vehicle should access, in what order, when to arrive, and when to complete the inspection.

[0092] Specifically, the initial inspection plan can include task package allocation results, access order, estimated arrival time, estimated completion time, navigation route, and dispatch instructions. In other words, by generating an inspection plan through a multi-vehicle route planning model, it is possible to map from a "risk-driven task set" to an "executable multi-vehicle inspection route," enabling the dispatch center to issue specific dispatch instructions to the driver or vehicle terminal based on the model output.

[0093] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may further include: acquiring inspection information of each inspection vehicle in real time, and performing state freezing processing on some inspection task packages based on the inspection information.

[0094] Specifically, inspection information can include the GPS location information of the inspection vehicle, the currently executing task, the remaining operation time, historical completed tasks, and the current path progress. Status freeze processing refers to marking completed task packages and currently executing task packages as locked. Completed task packages will no longer participate in subsequent optimizations, while currently executing task packages will maintain their vehicle and current execution relationship unchanged.

[0095] By freezing the status, it is possible to ensure that only the remaining tasks that have not yet been executed are performed, without changing the inspection behavior that has actually occurred or is in progress, thereby avoiding frequent changes to issued instructions that would affect the continuity of operations.

[0096] In some embodiments, the multi-vehicle route planning model can also select heuristic optimization methods, exact solution methods, or learning-based solution methods to solve the problem based on the road network scale and real-time requirements, so as to generate an inspection plan that meets the constraints.

[0097] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may further include: receiving dynamic monitoring data; dynamically updating the unlocked inspection task packages and aggregate weights based on the dynamic monitoring data to obtain updated task packages and updated aggregate weights; inputting the newly added task packages, updated task packages, and corresponding aggregate weights into a multi-vehicle path planning model to perform local re-optimization on the remaining inspection routes; and outputting an updated inspection plan based on the results of the local re-optimization to update the initial inspection plan.

[0098] The principle of local re-optimization in this application is: while keeping the locked task packages unchanged, only the allocation relationship and access order of the inspection task packages that have not yet been executed are adjusted, so as to quickly obtain a new inspection plan with a lower computational cost. Unlocked task packages can refer to task packages that have not yet been executed.

[0099] Because road defect reporting, traffic conditions, weather conditions, and vehicle progress may all change during actual inspections, the initial inspection plan can be adjusted in real time. Specifically, this application achieves closed-loop control of "initial planning - real-time updating - incremental adjustment" through a periodic dynamic refresh and local re-optimization mechanism. That is, the inspection task package and aggregate weights are updated online, and the initial inspection plan is partially modified.

[0100] In some embodiments, dynamic updates and local re-optimization can be triggered according to a preset refresh cycle. The preset refresh cycle can be 10 minutes, 15 minutes, or 30 minutes. In other embodiments, the preset refresh cycle can also be triggered by events such as sudden fault reports, severe congestion alarms, vehicle malfunctions, or task deviations exceeding a threshold.

[0101] In some embodiments, dynamic monitoring data may include disease reporting information, real-time traffic congestion information, weather and environmental data, and inspection execution feedback data.

[0102] In some embodiments, the updated comprehensive vulnerability index Ftotal′ can be recalculated for road segments that have not yet been inspected, and the updated inspection decision weight Wtask′ can be obtained by referring to the above method for obtaining inspection decision weights.

[0103] It should be noted that the updated inspection decision weight Wtask′ and the inspection decision weight Wtask can belong to the same weight system and be dynamically updated at different times; correspondingly, the updated aggregate weight can be represented as Wzone′, which represents the latest inspection benefit value of each unexecuted task package at the current refresh time.

[0104] Furthermore, if a new road segment with sudden defects appears, it can be re-aggregated with its adjacent road segments to generate a new task package or merged into an existing task package. Simultaneously, its updated aggregation weight Wzone′ is calculated to form a new pool of tasks to be planned. The calculation method for the aggregation weight Wzone′ can be found in the relevant description of aggregation weight calculation in this application above, and will not be repeated here.

[0105] In some embodiments, local re-optimization of the remaining inspection path may include: traversing any two inspection task packages in the remaining inspection path, obtaining several feasible insertion positions between any two inspection task packages that satisfy the constraints; calculating the change in the objective function of the feasible insertion positions, and selecting the feasible insertion position with the smallest change in the objective function to insert the new task package.

[0106] Constraints may include vehicle remaining time constraints, time window constraints, and accessibility constraints.

[0107] Specifically, for a new task package u, iterate through any two adjacent task packages p and q in the existing remaining paths of each vehicle, and try to insert the new task package u between p and q; if the vehicle's remaining time constraint, time window constraint, and accessibility constraint are still satisfied after insertion, then calculate the change in the objective function caused by inserting the task package.

[0108] In some embodiments, the change in the objective function can be expressed as: Where cpu represents the travel cost from inspection task package p to new task package u, cuq represents the travel cost from new task package u to inspection task package q; cpq represents the cost of traveling directly from inspection task package p to task package q; ΔT represents the additional delay on the vehicle's remaining path after inserting the new task package; W zoneu This represents the aggregate weight of the newly added task package; α is the first adjustment coefficient, β is the second adjustment coefficient, and γ is the third adjustment coefficient.

[0109] From all feasible insertion positions that satisfy the constraints, select the position that minimizes the change ΔJ as the optimal insertion point to enable the rapid integration of new high-priority tasks.

[0110] In some embodiments, the intelligent inspection route optimization method driven by multiple weights for urban roads may further include: partially exchanging unexecuted task packages in different vehicle paths, wherein the partial exchange may at least include exchanging the access order of two unexecuted task packages within the same inspection path, and / or transferring unexecuted task packages from one vehicle path to another vehicle path, and / or exchanging unexecuted task packages between two vehicles. Newly added task packages and updated task packages may also be exchanged.

[0111] Each partial swap allows for re-verification of the time window constraints, vehicle resource constraints, and objective function value after the swap. If the objective function is improved after the swap, the swap result is retained; otherwise, the original path scheme is restored. By repeatedly executing partial swaps, the overall inspection plan's benefits and execution rationality can be gradually improved.

[0112] After completing the local re-optimization, the new vehicle task allocation results, task package access order, estimated arrival time and route navigation information are output, which is the updated inspection plan, and sent to the driver terminal or vehicle equipment in real time.

[0113] Therefore, the remaining inspection tasks can be dynamically adjusted without disrupting the tasks already executed, achieving closed-loop control, real-time response, and multi-vehicle collaborative optimization during the inspection process.

[0114] This application establishes a dynamic closed-loop scheduling mechanism based on task aggregation and incremental re-optimization, which improves the continuity and emergency response capability of multi-vehicle inspection.

[0115] This application aggregates spatially continuous road segments with similar weights and compatible time windows into highly cohesive inspection task packages, and generates initial inspection routes based on a multi-vehicle path planning model with time window constraints. During the inspection execution process, unexecuted inspection task packages are adjusted in a rolling manner through state freezing, dynamic weight updates, task pool refresh, and local re-optimization mechanisms.

[0116] Therefore, without disrupting existing or ongoing tasks, it can quickly respond to sudden reports of malfunctions, changes in traffic conditions, and fluctuations in vehicle resources, shortening the replanning time after emergency tasks are inserted, and improving the stability, flexibility, and overall execution efficiency of multi-vehicle collaborative inspections.

[0117] In this application, the disease evolution chain is first obtained based on the spatiotemporal matching algorithm, and then the basic vulnerability characteristics are obtained from the disease evolution chain to calculate the inspection decision weight. Based on the inspection decision weight, the road segments to be inspected are aggregated into inspection task packages and corresponding aggregation weights. Finally, the initial inspection plan is output based on the inspection task packages and corresponding aggregation weights through the multi-vehicle path planning model.

[0118] In this context, by acquiring time-series-based disease evolution chains, it is possible to extract disease recurrence patterns from historical maintenance records, enabling inspection resources to be prioritized for high-risk road sections. This improves the probability of discovering high-risk disease points and the scientific rigor of inspection plans. Furthermore, aggregating road sections to be inspected into inspection task packages based on inspection decision weights allows subsequent inspection plans to consider risk priority, traffic conditions, and operational timeliness, thereby improving the coverage efficiency of high-value inspection tasks. Simultaneously, compared to solving discrete road sections one by one, this application reduces the number of path planning nodes and the scale of combinatorial optimization, thus improving the solution efficiency and engineering practicality of multi-vehicle path planning models in large-scale urban road networks.

[0119] Those skilled in the art will understand that all or part of the features / steps of the above-described method embodiments can be implemented by methods, data processing systems, or computer programs. These features may be implemented without hardware, entirely in software, or in a combination of hardware and software. The aforementioned computer program may be stored in one or more computer-readable storage media. When the computer program is executed (e.g., by a processor), it performs the steps of the above-described embodiments of the intelligent inspection route optimization method for multi-weighted urban roads.

[0120] The aforementioned storage media capable of storing program code include: static hard disks, solid-state hard disks, random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), optical storage devices, magnetic storage devices, flash memory, magnetic disks or optical disks, and / or combinations of the above devices, that is, they can be implemented by any type of volatile or non-volatile storage devices or combinations thereof.

[0121] This application also provides a processing device embodiment, including one or more processors and a memory; wherein the memory is used to store one or more computer programs, and the one or more processors are used to execute the one or more computer programs stored in the memory, so that the processors execute the features / steps of the above-described embodiment of the intelligent inspection route optimization method for multi-weighted urban roads.

[0122] The above description is merely a preferred embodiment of this application. Those skilled in the art will understand that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this application. Furthermore, under the teachings of this application, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of this application. Therefore, this application is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this application.

Claims

1. A method for optimizing intelligent inspection routes for urban roads based on multi-weighted driving principles, characterized in that, include: Several initial disease records are obtained, and the initial disease records are associated using a spatiotemporal matching algorithm to generate a disease evolution chain based on time sequence. Basic vulnerability characteristics are obtained from the disease evolution chain. Based on the basic vulnerability characteristics, internal vulnerability score and external influence score are calculated. The internal vulnerability score and external influence score are weighted to obtain a comprehensive vulnerability index. The comprehensive vulnerability index is mapped to an inspection decision weight. Based on the inspection decision weight, the road section to be inspected is aggregated to obtain an inspection task package and the corresponding aggregation weight. A multi-vehicle route planning model is constructed. The inspection task package and the aggregation weight are input into the multi-vehicle route planning model to obtain an initial inspection plan. The multi-vehicle route planning model uses the selection of the inspection task package with the largest aggregation weight and the minimization of inspection cost and inspection time as objective functions. The step of using a spatiotemporal matching algorithm to associate the initial disease records and generate a disease evolution chain includes: Choose one of the aforementioned initial disease records as a seed record, and search within a preset road segment range based on the seed record to obtain multiple candidate records, wherein the candidate records include initial disease records and maintenance records in subsequent time steps; Spatial and temporal consistency matching is performed on the candidate records. The candidate records that pass the spatial and temporal consistency matching are associated with the seed records to obtain the repair nodes and recurrence nodes. The discovery nodes, upgrade nodes, repair nodes and recurrence nodes corresponding to the same disease location are sequentially connected in time order to obtain the disease evolution chain. The process of aggregating inspection tasks based on the inspection decision weights to obtain inspection task packages and corresponding aggregation weights includes: Based on the road segment to be inspected, a road segment map model is constructed. According to preset conditions, each adjacent road segment in the road segment map model is judged and summarized to obtain a candidate aggregation set. The preset conditions include at least the difference between the inspection decision weights of adjacent road segments is not greater than a first preset threshold. The candidate aggregation set is subjected to structured encoding to obtain structured input data. The structured input data is aggregated and judged using a large language model to output the task package label corresponding to each road segment to be inspected. Based on the task package label, multiple road segments to be inspected are merged into the inspection task package. At the same time, the aggregation weight of the inspection task package is calculated. The aggregation weight is used to characterize the node reward value of the vehicle executing the inspection task package.

2. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 1, characterized in that, The intelligent inspection route optimization method driven by multiple weights for urban roads further includes: acquiring the inspection information of each inspection vehicle in real time, and performing state freezing processing on some of the inspection task packages based on the inspection information.

3. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 1, characterized in that, The intelligent inspection route optimization method driven by multiple weights for urban roads further includes: receiving dynamic monitoring data; dynamically updating the unlocked inspection task packages and the aggregated weights based on the dynamic monitoring data to obtain updated task packages and updated aggregated weights; inputting the newly added task packages, the updated task packages, and the corresponding aggregated weights into the multi-vehicle path planning model to perform local re-optimization on the remaining inspection routes; and outputting an updated inspection plan based on the results of the local re-optimization to update the initial inspection plan.

4. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 3, characterized in that, The local re-optimization of the remaining inspection path includes: traversing any two inspection task packages in the remaining inspection path, obtaining several feasible insertion positions between any two inspection task packages that satisfy the constraints; calculating the change in the objective function of the feasible insertion positions, and selecting the feasible insertion position with the smallest change in the objective function to insert the new task package.

5. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 3, characterized in that, The intelligent inspection route optimization method driven by multiple weights for urban roads further includes: partially exchanging unexecuted task packages in different vehicle paths, wherein the partial exchange includes at least exchanging the access order of two unexecuted task packages within the same inspection path, and / or transferring unexecuted task packages in one vehicle path to another vehicle path, and / or exchanging unexecuted task packages between two vehicles.

6. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 1, characterized in that, The aggregation weights are calculated as follows: Among them, ΣWtask i η1 is the cumulative value of the inspection decision weights of all the road sections to be inspected within the inspection task package; Uzone is the timeliness urgency of the inspection task package; Czone is the spatial compactness of the inspection task package; Dzone is the additional detour cost or dispersion of the inspection task package; η1 is the first preset weight coefficient, η2 is the second preset weight coefficient, η3 is the third preset weight coefficient, and η4 is the fourth preset weight coefficient.

7. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 4, characterized in that, The objective function is expressed by the following formula: Among them, W zone_p y represents the aggregate weight of the inspection task package p; p Indicates whether the inspection task package p is executed; cpq represents the travel cost from inspection task package p to inspection task package q; T k The total inspection time of vehicle k is represented by λ1, λ2 is the first preset adjustment coefficient, λ3 is the second preset adjustment coefficient, and the decision variable x(k, p, q) is used to indicate whether vehicle k travels from inspection task package p to inspection task package q.

8. The intelligent inspection route optimization method for urban roads driven by multiple weights according to claim 7, characterized in that, The change in the objective function is expressed as follows: Where cpu represents the travel cost from the inspection task package p to the new task package u, cuq represents the travel cost from the new task package u to the inspection task package q; cpq represents the travel cost from the inspection task package p directly to the inspection task package q; ΔT represents the additional delay on the vehicle's remaining path after inserting the new task package; W zoneu The aggregate weight of the newly added task package is represented by α, β, and γ.

Images